What You Practice, You Become:

Aristotle’s Warning About Delegated Thinking

There’s a passage in the Nicomachean Ethics where Aristotle says something that sounds almost too obvious to be worth saying: we become just by doing just acts, courageous by doing courageous acts, temperate by doing temperate acts. Of course we do. It’s the corollary that actually bites as we also become unjust by doing unjust acts, cowardly by avoiding situations that require courage, intemperate by repeatedly giving in to impulse. Character isn’t what you are, it’s what you keep doing.

This is Aristotle’s doctrine of habituation, ethismos, and it sits at the center of his entire ethical framework. The soul, he argues in Book II, acquires its virtues the same way we acquire craft skills: through exercise and repetition until a stable disposition, a hexis, forms. Once formed, that disposition is what makes you the kind of person who responds to situations in characteristic ways. A courageous person doesn’t have to white-knuckle it through every scary situation; courage has become, through practice, part of how they’re structured.

What’s important here, and what often gets lost in casual summaries of Aristotle, is that this isn’t just about moral virtue in the narrow sense. It applies equally to practical reasoning, what Aristotle calls phronesis, usually translated as practical wisdom. Phronesis is, for him, the master virtue. It’s the capacity to deliberate well about what to do, to perceive the morally relevant features of a situation, to weigh competing goods without a formula to fall back on. It’s the thing that allows virtue to actually function in the messy, particular circumstances of real life. Phronesis is also, crucially, something you can be bad at, and something you can get worse at by not using it.

This is relevant to how we talk about AI because if you consistently delegate your practical reasoning, you may not be losing some abstract essence of yourself. You may be losing the actual cognitive capacity to reason and choose well. The skill might atrophy.

There’s a body of research on cognitive offloading and the conditions under which it helps or hinders learning and capability development, though the picture is genuinely complicated and I won’t pretend the literature is settled. What Aristotle gives us is a mechanism for thinking about it: habituation works in both directions. The disposition forms through repetition, and it weakens through disuse. If the repeated experience is “encounter a decision, hand it to the AI, accept or lightly modify the output,” then that’s what you’re practicing.

What you’re not practicing is sitting with the difficulty of a problem, tolerating the discomfort of not knowing, generating options from your own resources, making a decision and living with it. Those are the micro-exercises through which phronesis develops. Or fails to. There’s a version of the counterargument that’s worth taking seriously: we’ve always used external cognitive tools. Writing itself is a form of cognitive offloading. Socrates famously worried about it in the Phaedrus, arguing that writing would weaken memory rather than aid it. We use calendars so we don’t have to remember dates, calculators so we don’t have to do arithmetic, GPS so we don’t have to navigate. And we’re fine, more or less.

But there are a few disanalogies that seem worth reflection. Most cognitive tools we use are transparent, you know when you’re using a calculator, and using it doesn’t feel like reasoning. AI assistance, particularly conversational AI that produces fluent, confident, contextually appropriate output, can blur that line in a way a calculator doesn’t. The output feels like your thinking process made more efficient, not like a separate tool doing a separate thing. That blurring might matter for what gets practiced and what doesn’t.

More importantly: arithmetic and calendar management are not phronesis. They’re not the capacities that Aristotle would put at the center of living well. The practical reasoning required in figuring out what to do in situations where the answer isn’t obvious, where values compete, where you have to decide who to be, is different in kind. Outsourcing the calculation of a tip is not the same thing as outsourcing the deliberation about a difficult conversation with a colleague, or how to handle a situation where you’ve been treated unfairly.

If you hand those to AI regularly, not for information, but for the actual deliberation, Aristotle’s framework suggests you should expect something to happen to the relevant capacity over time. Whether it does in practice, and at what threshold, isn’t something I can tell you but it’s a live question.

Aristotle is clear that virtue is not the same as having correct beliefs about what virtue requires. You can know perfectly well that courage is admirable and still be a coward, because knowledge alone doesn’t restructure the soul. What restructures the soul is doing. This matters because a person could theoretically engage very thoughtfully with AI-assisted reasoning, reviewing the output critically, pushing back, considering alternatives, and still, if the actual generative work is consistently elsewhere, not be practicing what they think they’re practicing.

The form of engagement matters, not just the content. None of this is an argument for refusing AI tools. That would be silly. It that there’s a difference between using AI to gather information faster, check your reasoning, expose you to framings you hadn’t considered, and using it to do the reasoning while you supervise. The first might be fine or even good for developing practical wisdom. The second, on Aristotelian grounds, is worth being cautious about.

Character, for Aristotle, is not something you have or don’t have. It’s something you’re always in the process of making, through every choice about how to meet the situation in front of you. The question isn’t whether AI changes who you are in some deep metaphysical sense. The question is whether the habits you’re forming with these tools are the habits of someone who can think well without them, and whether that matters to you.

In my opinion, it probably should.